the global digital divide revisited (2000-2004)

25
1 The Global Digital Divide Revisited (2000-2004) Maria Lurdes Castro Martins 1 NIMA- University of Minho Antonio Rodríguez Andrés Institut for Folkesundhed (Institute of Public Health) - Aarhus This version: October 2009 Abstract This paper examines the determinants of the “digital divide” using a macro panel data set of 188 countries over the period 1990-2004. Prior research shows that the “digital divide” relates to economic development, education, regulatory environment, internet costs, enforcement, personal computers, spoken language, and digital communications infrastructure, this paper also demonstrates that differences in the distribution of income (inequality) within countries can play a vital role when explaining the “digital divide”. Moreover, this effect can be so strong that, according to our results, the negative inequality effect may tend to exceed a possible positive direct income effect on Internet and personal computers diffusion processes. As a consequence, a reduction in income disparities can be essential to promote ICTs diffusion. Our results also suggest that an increase in the public investment in human capital as well as an effort in the promotion of markets openness can considerably reduce the “digital divide”. Keywords: Digital divide; Internet; Technological diffusion JEL classifications: L50; L96; O30 1 Corresponding author: NIMA – Núcleo de Investigação em Microeconomía Aplicada. School of Economics and Management. University of Minho, 4710-057 Braga, Portugal. Tel.: +35 1 253 604 533; Fax: +35 1 253 601 380; E-mail address: [email protected] .

Upload: others

Post on 17-Mar-2022

2 views

Category:

Documents


0 download

TRANSCRIPT

1

The Global Digital Divide Revisited (2000-2004)

Maria Lurdes Castro Martins1

NIMA- University of Minho

Antonio Rodríguez Andrés

Institut for Folkesundhed (Institute of Public Health) - Aarhus

This version: October 2009

Abstract This paper examines the determinants of the “digital divide” using a macro panel data set

of 188 countries over the period 1990-2004. Prior research shows that the “digital divide”

relates to economic development, education, regulatory environment, internet costs,

enforcement, personal computers, spoken language, and digital communications

infrastructure, this paper also demonstrates that differences in the distribution of income

(inequality) within countries can play a vital role when explaining the “digital divide”.

Moreover, this effect can be so strong that, according to our results, the negative inequality

effect may tend to exceed a possible positive direct income effect on Internet and personal

computers diffusion processes. As a consequence, a reduction in income disparities can be

essential to promote ICTs diffusion. Our results also suggest that an increase in the public

investment in human capital as well as an effort in the promotion of markets openness can

considerably reduce the “digital divide”.

Keywords: Digital divide; Internet; Technological diffusion JEL classifications: L50; L96; O30

1Corresponding author: NIMA – Núcleo de Investigação em Microeconomía Aplicada. School of Economics and Management. University of Minho, 4710-057 Braga, Portugal. Tel.: +35 1 253 604 533; Fax: +35 1 253 601 380; E-mail address: [email protected].

2

1. Introduction

Digital divide is a new phenomenon emerging with the development of the

information and communication technologies (henceforth, ICTs). The global economy is

being driven by greater integration of global markets and the spectacular growth of the

ICTs. The widespread use and implementation of ICTs has increased the world’s potential

for dissemination of knowledge and information. As a result, a positive sense has emerged

concerning the uses and potential benefits from the continued growth of the ICTs. ICT

sectors have been growing faster than non ICT sectors. Indeed, ICT services have been

growing even faster, particularly computers and related services.

The United Nations Development Report (UNDP) warned that the gains in

productivity produced by the new technology may widen differences in economic growth

between the most affluent nations and those that lack the skills, resources, and

infrastructure to invest in the information society (UNDP, 1999). Thus, poorer societies

might be marginalized at the periphery of the communication networks (Norris, 2002).

Indeed, there are important differences in the degree of diffusion or adoption across

countries. The gap between developed and developing countries is increasing over time

(Andrés et al., 2007).

Previous formal literature on the diffusion of ICTs has stressed that disparities in

ICTs diffusion may have an important role in the diffusion of knowledge, levels of

political engagement, as well as on economic growth (Norris, 2001; Steinmueller, 2001;

Brynjolfsson & Hitt, 2003; Wallsten, 2005). As a result, governments of developing

countries have become aware of how decisive a political strategy to eliminate the above-

mentioned effects can be, and have tried to catch up with more developed countries.

3

A better understanding of all aspects of digital divide is essential in order to be able

to implement adequate policy formulations as documented by international organizations

(See, for instance, UNCTAD,2005; World Bank, 2004). This paper aims to analyse the

permanence of a “digital divide” by considering a cluster of information networks that

include personal computers (PCs) and the Internet. The processes involved in diffusion,

emerging trends, and their magnitude can be important inputs for the design and

implementation of public policies in developing countries.

In 2004, thirty percent of the world’s population had 66% of the world’s GDP, 64%

of the world’s PCs, and they represented 58% of the world Internet subscribers and 75% of

Broadband users. Despite the high level of these disparities, it is interesting to note that the

Internet is experiencing a change in its trend toward inequality. In 1997, 93% of Internet

subscribers were concentrated among only a fifth of the world’s people (Kiiski & Pohjola,

2002).

In this paper, we focus on the digital divide. In this paper, we make use of a unique

dataset to study the determinants of digital divide for a large set of countries for the period

1990-2004. We use a dataset that covers more countries and years that earlier studies. Our

analysis includes both developed and developing countries. This paper addresses three

main questions:

1. Is there an international “digital divide” in the processes involved in providing

access to PCs and the Internet? Is the “digital divide” decreasing or increasing? Is it

seriously affected by each country income inequality?

2. Is regulation affecting the diffusion process? Is education a barrier or an enhancer

of diffusion?

4

3. What are the main determinants of the observed disparities? How can governments

promote public policies aimed at reversing the expansion lag in information

networks?

This study is very much in the same vein as that of Chinn & Fairlie (2007) focusing on

PCs, the Internet and international disparities, and uses a similar methodology. However,

this analysis go further by taking into account the specific role income inequalities within

countries can play in explaining ICTs diffusion disparities across countries. In addition,

this research adds new knowledge to the existing empirical studies by assessing whether

underdeveloped countries can really catch up with the more developed countries in terms

of access to the Internet. In addition, it considers several alternative statistical

measurements in order to study disparities in the world’s “digital divide”; breaking it down

into different components, and identifying the main determinants. To our knowledge, there

have been no empirical studies that address the same general question concerning the

effects of ICTs on the growth and development of a nation in this way. Further, we

conduct our empirical analysis, in a panel data framework, using standard panel data

techniques.

The results show that income and educational asymmetries can be important

factors affecting diffusion inequalities in the cross-country information clusters studied. In

addition, each country’s openness to trade can also play an important role explaining the

observed disparities. Finally, a crucial element explaining the persistence of the “digital

divide” can emerge from the asymmetrical distribution of income within countries.

The remainder of this paper is organized as follows: Section 2 discusses the

variables that are included in the baseline specifications. Our model, estimation strategy

5

and empirical results are presented in Section 3. Section 4 presents a series of robustness

checks and conclusions are drawn in Section 5. Summary statistics, correlation matrices

and regression results are included in the Appendix.

2. Variables and discussion

2.1 Interest variables

The digital divide is defined as the gap between those with a permanent, effective

access to new information and communication technologies (ICTs) and those with none

(e.g. Hoffman & Novak, 2000; Fairlie, 2004; Andonova, 2006; Chinn & Fairlie, 2007).

According to this definition, this gap can be witnessed both at the national level, between

different social groups, and internationally, between different countries. The present study

deals with the last one.

In attempting to measure the size and evolution of the international “digital divide”

a large set of variables have been tested as determinants of PCs and Internet diffusion.

Over the past ten years literature has endeavored to uncover economic, social, and political

factors that aided or hindered the divergence of ICTs diffusion rates across countries.

A substancial body of the literature has examined the impact of differences in

income, human capital, legal environment, and telecommunications infrastructures on

ICTs adoption (e.g. Harggitai, 1999; Quibria et al., 2000; Dasgupta et al., 2001; Oxley &

Yeung, 2001; Robison & Crenshaw, 2002; Kiiski & Pohjola, 2002; Bellock and

Dimitrova, 2003; Wallsten, 2005; Chinn & Fairlie, 2007). A more limited number of

studies have looked at the role of inequality across countries in influencing the

international “digital divide” (e.g. Harggitai, 1999; Kiiski & Pohjola, 2002).

6

The relationship between GDP and ICTs diffusion is well documented in literature.

For example, Harggitai (1999), Quibria et al (2000), Kiiski & Pohjola (2002), Bellock &

Dimitrova (2003), and Chinn & Fairlie (2007) all have shown that GDP is a large

determiner of Internet access. International disparities in per capita income help to explain

the gap in computer and Internet use. But this is not the only important factor afecting the

global “digital divide”.

Bellock & Dimitrova (2003) showed that increasing civil liberties have also a

positive and significant effect on the Internet diffusion process. Robison & Crenshaw

(2002) that development level, political openness, mass education, and the size of tertiary

sector are the most significant determiners of Internet penetration. Also, Kiiski & Pohjola

(2002) find that education can be an important factor in Internet diffusion when

developing countries are included in the sample.

Oxley & Yeung (2001) demonstrated that Internet hosts penetration is positively related

with telecoms infrastructures, rule of law, and credit card use and negatively correlated

with telephone service costs. Quibria et al (2000) analysing PCs and Internet use per capita

find that GDP, education levels, and infrastructure are the most important drivers of these

ICTs diffusion. More recently, Chinn & Fairlie (2007) studied PCs and Internet use per

capita and find that GDP, telephone density and regulatory quality (pro-market policies)

are important determinants of these technologies.

Following the above mentioned surveyed results, in the present analysis we

considered that the process of diffusion of PCs and Internet is affected by: economic

7

indicators, human capital, the institutional and legal environment, and the development

level of the existing technological infrastructure.

In our study, the selected economic indicators are: telecommuniactions prices, the

level of income per capita (GDP), the Gini index of inequality in the distribution of income

for each of the 188 countries in the sample and the degree of international exposure of

national markets to international competition. As proxies for telecommunications prices

we used both, the price of a three minutes mobile call at peak rate, and the price of a three

minutes local call at peak rate (International Telecommunications Union (ITU) database).

Income per capita data comes from the Penn World Tables and the source for the Gini

index variable is the World Bank’s World Development Indicators (WDI) database. Total

trade as a percentage GDP is the variable employed to measure the degree of openness of

national markets.

The country’s educational level is assessed by the level of public expenditures in

education as a percentage of the country’s gross domestic product. Legal environment is

introduced in the model using data from the International country risk guide (PRS group),

more specifically from Table 3B: political risk points by component.

The existing technological infrastructure is captured by the existing main lines in

operation for each country in the sample and the percentage of urban population as this last

variable can help us to estimate possible cost differences in building the fixed

telecommunications infrastructure. Both variables come from the International

Telecommunication Unions (ITU) database.

3. EMPIRICAL STRATEGY

8

Technological innovations such as the ICTs are usually not immediately adopted

by all potential consumers. The formal literature has pointed out that individuals have a

different timing to innovate (Lefebvre & Lefebvre, 1996). It is possible to classify their

behaviour across five different categories: innovators, early adopters, early majority, late

majority, and laggards (Rogers, 2003). Based on the concept of rate of adoption, an

innovation diffusion process has been viewed as following a pattern over time that seems

like an S-shaped or sigmoid curve: first the innovation adoption rate is slow, after that it

starts experiencing a phase of rapid growth, after which it stabilizes, and eventually

declines. The sigmoid diffusion functions’ second derivatives are positive first and then

negative after the inflection point. The literature typically uses two kinds of models for

ICT diffusion (Griliches, 1957; Dixon, 1980; Pereira & Pernías-Cerrillo, 2005): Gompertz,

and logistic. Both represent S shaped diffusion paths.

Figure 1: S-shaped computers diffusion Source

The important feature of the Gompertz curve is that the diffusion goes faster at the

beginning but becomes slower over time. This leads to a relatively short period of rapid

expansion and to a relatively long period of gradual growth up to the maximal level. The

logistic curve is more symmetric - the growth rate is initially not as high as in the

9

Gompertz curve and it declines more gradually (see Jarne, Sanchez-Choliz, & Fatas-

Villafranca, 2007).

This S-shaped curves are also designated by “epidemic models” of technology

diffusion and they have been the major approach used in the innovations diffusion

literature since the middle of the twentieth century. The flow of new adopters of the

innovation is related to the stock of existing adopters. When this stock is small the risk of

“contagion” is low but as the stocks grows the risk of contagion increases. And as soon as

the stock reaches a level close to the total number of potential adopters the flow of new

adopters decreases. This approach major drawback is that they do not give an economic

explanation for the spread of the innovation, the diffusion process is exogenously given

and do not take into account individuals adoption choices.

Figure 2: S-shaped Internet diffusion process Source

In the curent study, we follow a well known literature alternative approach (e.g.

Kiiski & Pohjola, 2002; Madden et al., 2004; Gruber, 2005) that combines the constant

parameters of the diffusion model with other variables that introduce more flexibility into

the innovation diffusion process. The diffusion process is affected by both the location

variables and the growth variables. These variables depend on country specific effects as

10

well as on a set of continuous explicative varibles. Any particular observation for country i

and period t of the diffusion process can be modelled by the equation:

yi t= yi* /(1+exp(-ait-bitt)) (1)

where Y* is the number of potential adopters, and a and b are the above mentioned

location and growth parameters. According to this model

dyt/d t= byt (1-yt/y*) (2)

and d2yt/dt2 is positive until the inflection point after which it turns to be negative.

Following Kiiski & Pohjola (2002), and Chinn & Fairlie (2007), we use a

Gompertz model of technology diffusion to test the role of the Internet and PCs cross-

country diffusion, with the aim of assessing the evolution of the digital divide in more

recent years. This models start from equation (2) and assume that the number of potential

adopters is a linear function of variables such as per capita income (GDP), citizens

educational level, openness of the countries markets, prices of the technologies being

analysed, the size of telecommunications infrastrucure, and regulatory environment. To

identify the main determinants of internet adoption, we follow Estache et al. (2002), and

Caselli and Coleman (2001), and estimate the following reduced form:

ln(Yit/Yit-1) = β0 + β1* Pit + β2* GDPit + β3* Nit-1 + β4* I it + β5* Zit+ uit (3)

where Y is the penetration rate in country i at period t, the βs are parameters to be

estimated and uit is a zero-mean stochastic error structure. The explanatory variables

include a constant term, P is a price vector, the real GDP per capita (GDP), the lagged

value of the network size (Nt-1 ), an inequality measure (I), and a vector of socio-economic

variables (Z) that are expected to help to characterize each country location and growth

variables.

11

As economic inequality and income are related to each other, in this paper, we also

test the classical Kuznets hypothesis (Kuznets, 1955; Lewis, 1954). According to this

hypothesis, inequality rises with income at low levels but falls once income reaches a

critical level. Thus, the second derivate is negative, by regressing the following equation:

I it = µ0 + µ1 GDPit+ µ2 GDP2it + ηit (4)

We validate the Kuznets hypothesis if the coefficient of the linear term of this

equation has a positive sign (µ1), while the coefficient estimate of the quadratic term has a

negative value (µ2). If we substitute I it as defined in equation (4) into equation (3) we

obtain:

ln(Yit/Yit-1) = π0 + β1Pit+ π1 GDPit+ π3 GDP2it +β3Nit-1+β5Zit+uit (5)

where π0 = α + β4µ0, π1 = β2+ β4µ1 and π3 = β4µ2. We fit this model to a panel data set for

188 countries for the period 1990 through 2004.

4. THE DATA AND EMPIRICAL RESULTS

Data are gatherered from several sources. Complete data were available for

countries for the years 1990 through 2004. The countries included in the full sample are

listed in the Appendix. Descriptive statistics of the full sample are reported in Table 2. For

the full sample, the average number of computers…… (etc).

The individual effects are significant. All models reject the null hypothesis that

there is no correlation between individual effects and the explanatory variables, being the

prefereed specification the fixed effects model. Inspection of the inequality model

regression results enables us to validate the Kuznets hypothesis: the GDP coefficients are

positive (µ1 > 0) and statistically significant, documenting a positive impact of a GDP

12

increase on the inequality variablel, and the GDP2 estimated coefficient is negative (µ2 <

0) and statistically significant, stating that the previous effect is decreasing2.

Therefore, gathering information on the above mentioned results of the inequality

model and on the results of the diffusion models helps us explainig the negative sign of the

GDP coefficient and the positive sign of GDP2 in the diffusion model results. They both

reinforce the importance of the negative effect of inequality (β4 < 0) on diffusion processes

such as the Internet or personal computers. Moreover, it seems that, according to these

results, the negative inequality effect tend to exceed a possible positive direct income

effect (β2>0).

In line with previous studies, we also find evidence that there is a positive and

statistically significant impact of the degree of openness of markets on Internet and

personal computers diffusion. As we would expect, the degree of international exposure to

competition seems to be an important driver of ICTs diffusion.

Similar to the results already surveyed, we also show that education can help ICTs

diffusion rates. According to Table 3 and Table 4 results, education has a positive and

statistically significant impact on the adoption rate level attained, consistent with the

findings of other authors. Less evident is the negative impact of the telecommunications

infrastructure, when measured by the number of fixed mainlines, on the diffusion rates

being studied. The existing technological infrastructure captured by the existing main lines

in operation for each country in the sample seem to play a negative role in adoption rate.

2 We are here excluding from analysis Model 1 estimates. These are pooled regression results that are valid only on the strong assumption that there are no country specific effects. And as can be viewed in table 5 we reject the null hypothesis of the F-test.

13

Also, the percentage of urban population seems to have not a significant role in the rate of

adoption.

5. Conclusions

In this paper, we tried to assess the factors which determine the diffusion of

Internet and personal computers across countries placing a particular emphasis on the role

of inequality. Our results show that many of the differences observed in the use of

information technologies frequently designated by “digital divide” can emerge from

differences in the distribution of income within countries (inequality). Consequently, in

order to catch up with more advanced countries in ICTs expansion may require less

developed countries change public policy towards reducing income inequalities at the

national level.

Secondly, education and the degree of openness of the national markets are

important determiners of information technologies diffusion. Governments may consider

influencing these factors if the ICTs diffusion is one of their priorities.

Testing other factors that can contribute to the observed differences in the level of

information technologies diffusion across countries is the next step of this work. New

variables will be introduced and tested in order to validate current results. This analysis is

our starting point to comprehend the role of inequality and other economic, social and

political factors in explaining the global “digital divide”. Further work is needed to verify

the validity of the current conclusions.

In this study, there seems to be no clear difference between the results for the

Internet penetration rate gaps and those for the computer penetration rate gaps. As a

consequence public policies seem to be able to successfully attain a reduction in each of

14

the analysed gaps by promoting an increase in educational achievement, or contributing to

the openness of its markets, as well as by making an effort to reduce income disparities

among its population.

Acknowledgements

We would like to thank seminar participants at the 19th European regional conference of

the International Telecommunications Society in Rome, Italy, 18-20 September 2008 for

comments and suggestions. Antonio Rodríguez Andrés acknowledges financial support

from the Mideplan Scientific Initiative Regional Science.

15

References

Andonova, V. (2006). Mobile phones, the internet and the institutional environment. Telecommunications Policy, 30(1), 29-45. L. Andres, D. Cuberes, M.A. Diouf, and T. Serebrisky (2007). Diffusion of the Internet: A Cross-Country Analysis. World, 2007.

Antonelli, C. (2003). The digital divide: understanding the economics of new information and communication technology in the global economy. Information Economics and Policy, 15(2), 173-199. Bellock, R., & Dimitrova, D. (2003). An explanatory model of inter-country internet diffusion. Telecommunications Policy, 27(3-4), 237-252. Brynjolfsson, E., & Hitt, L. (2003). Computing productivity: firm-level evidence. Review of Economics and Statistics, 85(4), 793-808. Caselli, F., & Coleman II, W.J. (2001). Cross-country technology diffusion: The case of computers. American Economic Review, vol. 91, No. 2, Papers and Proceedings of the Hundred Thirteenth Annual Meeting of the American Economic Association, May, 328-335. Chinn, M., & Fairlie, R. (2007). The determinants of the global digital divide: a cross country analysis of computer and Internet penetration. Oxford Economic Papers, 59(1), 16-44. Dasgupta, S.; Lall, S., & Wheeler, D. (2005). Policy reform, economic growth and the digital divide. Oxford Development Studies, 33(2), 229-243. Dixon, R. (1980). Hybrid corn revisited. Econometrica, 48(6), 1451-1462. Estache, A., Manacorda, M., & Valletti, T.M. (2002). Telecommunications, reform, access regulation, and internet adoption in Latin America. Economía, Volume 2, Spring. Fairlie, R. (2004). Race and the digital divide. Contributions to Economic Analysis & Policy, 3(1), Article 15. Goldfarb, A., & Prince, J. (2008). Internet adoption and usage patterns are different: implications for the digital divide. Information Economics and Policy, 20(1), 2-15. Griliches, Z. (1957). Hybrid corn: An exploration in the economics of technological change. Econometrica, 25(4), 501-522. Gruber, H. (2005). The Economics of Mobile Telecommunications. Cambridge University Press, Cambridge. Harggitai, E. (1999). Weaving the Western Web: Explaining differences in Internet connectivity among OECD countries. Telecommunications Policy, 23(10-11), 701-718.

16

Hoffman, D., & Novak, T. (2000). The growing digital divide: implications for an open research Agenda in Brynjolfsson, E.; Kahin, B. (eds.), Understanding the digital economy: data, tools and research; The MIT Press, Cambridge, pag. 245-260. International Telecommunications Union (2005). World Telecommunication Indicators/ICT indicators database. Geneva, Switzerland. Jarne, G., Sanchez-Choliz, J., Fatas-Villafranca, F. (2007). S-shaped curves in economic growth. A theoretical contribution and an application. Evolutionary and Institutional Economics Review, 3(2), 239-259.

Kiiski, S., & Pohjola, M. (2002). Cross-country diffusion of the Internet, Information Economics and Policy, 14(2), 297-310. Kuznets, S. (1955). Economic growth and income inequality. American Economic Review

45(1), 1-28. LaRose, R., Gregg, J.; Strover, S., Straubhaar, J., & Carpenter, S. (2007). Closing the rural broadband gap: Promoting the adoption of the internet in rural America. Telecommunications Policy, 31(6-7), 359-373. Lefebvre, E, & Lefebvre, L. (1996). Information and telecommunications technologies: the impact of their adoption on small and medium-sized enterprises. International Development Research Centre, Ottawa. Lewis, W.A (1954). Economic development with unlimited supplies of labour. Manchester

School 22(2), 139-191. Madden, G.; Coble-Neal, G., & Savage, S. (2004). United States internet penetration. Applied Economics Letters, 11(9), 529-532. Norris, P. (2001). Digital Divide?: Civic Engagement, Information Poverty and the Internet Worldwide, Cambridge University Press, Cambridge. Norris, P. (2002).The Global Divide: Information poverty and internet Access Worldwide, Mimeo, John F. Kennedy School of Government, Harvard University. Oxley, J., & Yeung, B. (2001). E-commerce and readiness institutional environment and international competitiveness, Journal of International Business Studies, 32(4), 705-723. Pereira, P. and Pernías-Cerrillo, J. (2005). The diffusion of cellular telephony in Portugal before UMTS: A time series approach, Working Papers Autoridade Portuguesa da Concorrência. Quibria, M.; Ahmed, S.; Tschang, T., & Reyes-Macasaquit, M. (2003). Digital divide: determinants and policies with special reference to Asia. Journal of Asian Economics, 13(6), 811-825. Robison, K., & Crenshaw, E. (2002) Post-industrial transformations and cyber-space: A cross national analysis of internet development. Social Science Research, 31(3), 334-363.

17

Rogers, E. (2003). Diffusion of Innovations, Free Press, New York. Steinmueller, W. (2001). ICTs and the possibilities of leapfrogging by developing countries. International Labour Review, 140(2), 193-210. Teulings, C., & Van Rens, T. (2008). Education, growth and income inequality. The Review of Economics and Statistics, 90(1), 89-104. Van Dijk, J. (2005). The Deepening Divide: Inequality in the Information Society, Sage Publications, Thousand Oaks CA. Wallsten, S. (2005). Regulation and internet use in developing countries. Economic Development and Cultural Change, 53(2), 501-523. UNCTAD. The Digital Divide Report: ICT Diffusion Index. United Nations Publications, 2005. UNDP. 1999. Human Development Report 1999. NY: UNDP/Oxford. United Nations. WIID2 database

18

Appendix

Table A1. List of countries

01. Afghanistan 48. Djibouti 95. Latvia 142. Samoa 02. Albania 49. Dominica 96. Lebanon 143. Sao Tome Principe 03. Algeria 50. Dominican Rep. 97. Lesotho 144. Saudi Arabia 04. Angola 51. Ecuador 98. Liberia 145. Senegal 05. Antigua and Barbuda 52. Egypt 99. Libya 146. Serbia and Montenegro 06. Argentina 53. El Salvador 100. Lithuania 147. Seychelles 07. Armenia 54. Eq. Guinea 101. Luxembourg 148. Sierra Leone 08. Australia 55. Eritrea 102. Macao 149. Singapore 09. Austria 56. Estonia 103. Macedonia 150. Slovak Republic 10. Azerbaijan 57. Ethiopia 104. Madagascar 151. Slovenia 11. Bahamas 58. Fiji 105. Malawi 152. Solomon Islands 12. Bahrain 59. Finland 106. Malaysia 153. Somalia 13. Bangladesh 60. France 107. Maldives 154. South Africa 14. Barbados 61. Gabon 108. Mali 155. Spain 15. Belarus 62. Gambia, The 109. Malta 156. Sri Lanka 16. Belgium 63. Georgia 110. Mauritania 157. St. Kitts & Nevis 17. Belize 64. Germany 111. Mauritius 158. St. Lucia 18. Benin 65. Ghana 112. Mexico 159. St.Vincent Gren 19. Bermuda 66. Greece 113. Micronesia, F.

Sts. 160. Sudan

20. Bhutan 67. Grenada 114. Moldova 161. Suriname 21. Bolivia 68. Guatemala 115. Mongolia 162. Swaziland 22. Bosnia and Herzeg 69. Guinea 116. Morocco 163. Sweden 23. Botswana 70. Guinea-Bissau 117. Mozambique 164. Switzerland 24. Brazil 71. Guyana 118. Namibia 165. Syria 25. Brunei 72. Haiti 119. Nepal 166. Taiwan 26. Bulgaria 73. Honduras 120. Netherlands 167. Tajikistan 27. Burkina Faso 74. Hong Kong 121. Netherlands Ants 168. Tanzania 28. Burundi 75. Hungary 122. New Zealand 169. Thailand 29. Cambodia 76. Iceland 123. Nicaragua 170. Togo 30. Cameroon 77. India 124. Niger 171. Tonga 31. Canada 78. Indonesia 125. Nigeria 172. Trinidad &Tobago 32. Cape Verde 79. Iran 126. Norway 173. Tunisia 33. Central African Rep. 80. Iraq 127. Oman 174. Turkey 34. Chad 81. Ireland 128. Pakistan 175. Turkmenistan 35. Chile 82. Israel 129. Palau 176. Uganda 36. China 83. Italy 130. Panama 177. Ukraine 37. Colombia 84. Jamaica 131. Papua New

Guinea 178. United Arab Emirates

38. Comoros 85. Japan 132. Paraguay 179. United Kingdom 39. Congo, Dem. Rep. 86. Jordan 133. Peru 180. United States 40. Congo, Republic of 87. Kazakhstan 134. Philippines 181. Uruguay 41. Costa Rica 88. Kenya 135. Poland 182. Uzbekistan 42. Cote d`Ivoire 89. Kiribati 136. Portugal 183. Vanuatu 43. Croatia 90. Korea, Dem. Rep. 137. Puerto Rico 184. Venezuela 44. Cuba 91. Korea, Republic of 138. Qatar 185. Vietnam 45. Cyprus 92. Kuwait 139. Romania 186. Yemen 46. Czech Republic 93. Kyrgyzstan 140. Russia 187. Zambia 47. Denmark 94. Laos 141. Rwanda 188. Zimbabwe

19

Table 1. Descriptive statistics Variables Observations Mean Standard

Deviation Minimum Maximum

Income inequality- Gini index 201 30.956 7.687 19.6 62.5

Real Income (p.c) (in 1,000 $) 1158 8.753 1.075 6.198 10.835

Openness 1158 90.425 55.896 2.142 462.926

Main fixed lines 1157 4 967 698 1.66e+07 3082 3.12e+08

Bureaucracy quality 785 2.560 1.0248 1 4

Urban population (%) 701 56.662 22.958 5.78 100

Education expenditures (%) 994 5.116 3.201 1.323 45.855

Price local calls 1164 1.447 13.201 0.000 230.088

Price mobile calls 1164 169.487 2 568.834 0.003 79 610.340

Personal computers adoption rate 999 -10.628 1.802 -16.219 -4.683

Internet adoption rate

1004 -11.092 1.784 -16.337 -6.365

20

Table 2. Correlation matrix of variables used in regressions

Note: * indicates statistical significance at the 5 percent level.

Internet adoption rate

PCs adoption

rate

Real Income

(p.c)

Education Expendi-tures (%)

Openness Main fixed

lines per …

Bureau- cracy

quality

Urban popula-tion (%)

Price local calls

Price mobile calls

Lag PCs

Internet adoption rate 1

PCs adoption rate 0.956* 1

Real Income (p.c) 0.030* 0.091* 1 Education expenditures (%) 0.301* 0.313* 0.062* 1

Openness 0.430* 0.454* 0.267* 0.111* 1

Main fixed lines -0.436* -0.419* 0.191* -0.098* -0.176* 1 Bureaucracy quality 0.026 0.077* 0.774* 0.163* 0.197* 0.2996* 1 Urban population (%) 0.011 0.084* 0.722* 0.046* 0.200* 0.085* 0.450* 1

Price local calls 0.010 -0.013 -0.047* -0.080* -0.0045 -0.023* -0.055* 0.005 1

Price mobile calls 0.010 -0.013 -0.036* -0.064* 0.0003 -0.016 -0.034* -0.003 0.751* 1

21

Table 3. Regression results. Dependent variable: Internet adoption rate

Explanatory variables Model (1)

Model (2)

Model (3)

Real Income (p.c) -10.366***

[1.244] -3.257***

[1.338] -22.146***

[7.459] Real Income (p.c) 2 0.612***

[0.071] 0.1876***

[0.079] 1.245***[0.411]

Openness 0.012***

[0.002] 0.101***

[0.002] -0.002

[0.006] Education expenditures (%) 0.113***

[0.042] 0.071***

[0.024] -0.017

[0.011] Main fixed lines -0.001***

[0.001] -0.001***

[0.001] 0.001*

[0.001] Urban population (%) 0.013**

[0.006] 0.006

[0.006] 0.007 [0.020]

Bureaucracy quality -0.805***

[0.158] -0.162

[0.141] 0.096 [0.168]

Price local calls 0.009*

[0.005] 0.001

[0.004] -0.001

[0.001] Price mobile calls 0.001

[0.001] 0.001*

[0.004] 0.001***[0.001]

Lag internet users 0.001***

[0.001] 0.001**

[0.001] 0.001 [0.001]

Constant 32.163***[5.359]

1.930

[5.673] 86.566***

[3.057] Observations 397 397 397

Individual effects F-test (p-value) (0.000)Individual effects vs Random effects Hausman test (p-value)

(0.000)No serial correlation Baltagi-Wu Ibi statistic 2.158 Notes: * significant at 10%; ** significant at 5%; *** significant at 1%. Robust standard errors in brackets Model (1) Regression with robust standard errors and analytical weights. Model (2) Prais-Winsten regression, heteroskedastic panels corrected standard errors. Model (3) Fixed effects with standard errors adjusted for panel clustering.

22

Table 4. Regression results. Dependent variable: Personal computers

Explanatory variables Model (1)

Model (2)

Model (3)

Real Income (p.c) -10.169***

[1.029] -4.642***

[1.307] -0.183

[3.861]

Real Income (p.c) 2 0.597***

[0.060] 0.2811***

[0.078] 0.010

[0.212] Openness 0.011***

[0.001] 0.0078***

[0.001] -0.0002 [0.002]

Education expenditures (%) 0.124***

[0.032] 0.0624***

[0.017] -0.0005 [0.005]

Main fixed lines -0.0001***

[0.001] -0.0001**

[0.001] 0.0001*

[0.001] Urban population (%) 0.0186***

[0.005] 0.0070 [0.006]

0.0030 [0.005]

Bureaucracy quality -0.686***

[0.116] -0.0977 [0.101]

-0.013

[0.032] Price local calls 0.010***

[0.002] 0.0019 [0.001]

-0.0014***

[0.001] Price mobile calls -0.0001

[0.001] 0.0001 [0.001]

-0.0001 [0.001]

Personal computers -lag 0.0001

[0.001] -0.0001**

[0.001] 0.0001***

[0.001] Constant 29.812

[4.342] 6.557

[5.434] -10.864 [17.454]

Observations 315 315 315

R-squared

Individual effects F-test (p-value)

(0.000) Fixed effects vs Random effects Hausman test (p-value)

(0.000) No Serial Correlation Baltagi-Wu Ibi statistic 2.329 Notes: * significant at 10%; ** significant at 5%; *** significant at 1% Model (1) Regression with robust standard errors and analytical weights. Model (2) Prais-Winsten regression, heteroskedastic panels corrected standard errors. Model (3) Fixed effects with standard errors adjusted for panel clustering.

23

Table 5. Regression results. Dependent variable: Gini inequality index

Explanatory variables Model (1)

Model (2)

Model (3)

Real Income (pc)

-56.269*

[3.0537]

11.434*** [2.1276]

53.446*** [2.4009]

Real Income (p.c) 2 2.580

[1.5719] -0.811***

[0.2145] -2.592***

[1.2483]

Constant 333.762***

[1.4812] -2.928***

[0.5901] -243.131***

[115.7832]

Observations 276 276 276

R-squared

Individual effects F-test (p-value) (0.000) Fixed effects vs Random effects Hausman test (p-value) (0.000)

No Serial Correlation Baltagi-Wu Ibi statistic 1.500 Notes: * significant at 10%; ** significant at 5%; *** significant at 1% Model (1) Regression with robust standard errors. Model (2) Fixed effects with ar(1) disturbances. Model (3) Fixed effects with standard errors adjusted for panel clustering.

24

Table 6

Variable names, definitions and sources

Variable Variable description Source

Income Real GDP per inhabitant (chain index)

Penn World Tables (PWT) 6.2, 2006

Openness Total trade as a % of GDP

Penn World Tables (PWT) 6.2, 2006

Inequality Gini index (WIID2.c high quality criteria)

UNU-WIDERWIID2.c high quality criteria

Main fixed lines per 1000 people??? Main fixed telephone lines in operation divided by population

International Telecommunication Union (ITU), 2005

Urban population (%) Urban population as a percentage of total population

International Telecommunication Union (ITU), 2005

Bureaucracy quality

Quality of bureaucracy International Country Risk Guide (PRS group)(Table 3B: political risk points by component)

Price local calls Price of a 3-minute fixed telephone local call (peak rate) ppp adjusted

International Telecommunication Union (ITU), 2005

Price mobile calls Mobile cellular - price of 3-minute local call (peak) ppp adjusted

International Telecommunication Union (ITU), World Telecommunications Report, 2005

Education Public expenditure on education (%)

UNESCO Education Database

Internet adoption rate Adapted from ITU world telecommunications indicators database 2005

Personal computers adoption rate Adapted from ITU world telecommunications indicators database 2005

Internet users Description per 100 International Telecommunication Union (ITU), World Telecommuncations Report, 2005

Personal computers Description per 100 International Telecommunication Union, World Telecommunications Report (ITU), 2005

25